In their keynote article, Dijkstra, Wahl, Buytenhuijs, van Halem, Al-jibouri, De Korte & Rekke (Reference Dijkstra, Wahl, Buytenhuijs, van Halem, Al-jibouri, de Korte and Rekké2018) propose a new model that aims to integrate and take care of the possible shortcomings of both the Revised Hierarchical Model (RHM, Kroll & Stewart, 1994) and Bilingual Interactive Activation model (BIA and several of its later versions, Dijkstra & Van Heuven, Reference Dijkstra, Van Heuven, Grainger and Jacobs1998). They begin their proposal by examining the objections/issues raised by Brysbaert and Duyck (Reference Brysbaert and Duyck2010) on RHM. It is well known that RHM is a developmental, production-based model which mostly predicted performance on translation-based tasks albeit with different predictions for translation recognition vs. production with regard to second language proficiency; BIA being a connectionist model of bilingual word recognition majorly emphasized on language non –selective selection and parallel language activation. Having been developed in and around Dutch (English as the second language), it took cognate status and orthographic similarity between words very seriously. Cognate status and orthographic similarity as factors won't count much as theoretical constructs around which a hypothesis could be developed if we look round the diverse types of orthographies and phonologies we find around the world.

Therefore, while Multilink aims high with many simplified assumptions that account for translation, naming and word recognition, it simply won't have any success when we talk of the complex mechanisms of psycholinguistic processes for bilinguals whose two languages do not share any orthography or even phonology (take Chinese–English or Hindi–English). Therefore, it is not evident how this model is going to be relevant for those researchers who are trying to understand the psycholinguistics and cognitive processes of bilinguals whose languages often belong to two different language families. Further, considering the fact that the model mainly explains visual word recognition, and reading is an acquired habit (unlike speaking and listening), it's also not clear how we understand bilingualism with illiterates. Recent work has shown fundamental differences between the literate and the illiterate brain in terms of neural processing and cognition (e.g., Skeide, Kumar, Mishra, Tripathi, Guleria, Singh, Eisner & Huettig, Reference Skeide, Kumar, Mishra, Tripathi, Guleria, Singh, Eisner and Huettig2017). The constraints within which Dijkstra et al. (Reference Dijkstra, Wahl, Buytenhuijs, van Halem, Al-jibouri, de Korte and Rekké2018) have developed their model is understandable, but it is time we talked seriously about the diversity of the bilingual mind in the cross-linguistic and cross-cultural context and built truly general models.

To give an example from another domain, the EZ model (Reichle, Pollatsek, Fisher & Rayner, Reference Reichle, Pollatsek, Fisher and Rayner1998) that accounts for eye movements in the context of sentence reading has now been extended to account for reading in Chinese (Rayner, Li & Pollatsek, Reference Rayner, Li and Pollatsek2007). EZ was originally a serial processing model developed to explain eye movement control during English reading. However, since English and Chinese orthographies are very different, attentional demands required to process them also vary in nature. The model now includes specific features to account for this. Similarly, in the context of bilingual language processing, an extension of the Multilink model to other languages will be welcome.

Dijkstra et al. (Reference Dijkstra, Wahl, Buytenhuijs, van Halem, Al-jibouri, de Korte and Rekké2018) also aspire to propose a control mechanism that will take care of the spurious activations that the bilingual mind entertains during language processing. The real conflict between the cognitive psychology and psycholinguistics of bilingualism lies in explaining control. While language non-selective activation and parallel language activation in multiple modalities has been shown, we have never successfully explained how the brain selects the task-relevant lexical word and avoids distractions. Green's (Reference Green1998) model proposed inhibition to account for language schema selection. Multilink assumes parallel inhibition in a connectionist framework. But does inhibition work similarly for non-selective activation in naming, translation and spoken word recognition? It is likely that individual differences with regard to core executive control abilities modulate language selection in these domains. The current version of the Multilink model does not explicitly refer to the cognitive (general) control mechanisms that regulate language activation. Also, are the control mechanisms similar for the language pairs that have cognates and orthographically similar words as opposed to those that don't have them? Currently, we know very little about general neural mechanisms that help us achieve fluency in language processing in the event of pervasive and continuous language non-selective activation.

Recent evidence on bilinguals’ language non-selective activation has come from the visual world paradigm (Tanenhaus, Spivey-Knowlton, Eberhard & Sedivy, Reference Tanenhaus, Spivey-Knowlton, Eberhard and Sedivy1995). For example, the BLINCS model (Shook & Marian, Reference Shook and Marian2013) explains the multi-modal nature of bilingual language processing with spoken words. It has been shown that bilinguals with language pairs that do not have cognates or any orthographic similarity engage in spontaneous translation during audio-visual speech processing (e.g., Mishra & Singh, Reference Mishra and Singh2016). It appears that language non-selective activation is a property of the bilingual mind and is not dependent much on cognate pairs found in the languages. Thus, the Multilink has limited usefulness to the vast number of studies that have shown consistent cross-linguistic activation on multiple levels when bilinguals engage in audio-visual speech processing. Spoken words also influence orthographic processing in such a paradigm (Mishra & Singh, Reference Mishra and Singh2014; Salverda & Tanenhaus, Reference Salverda and Tanenhaus2010). Although the Multilink model explains much of the authors' own data, it's important that cross-modal data related to non-selective activation and control is considered. Visual world studies have shown both form-based (bottom-up) and cross-linguistic semantic activation. Eventually, the executive control that is required to sort out languages has to consider the real nature of the bilingual interaction: that is, listening, looking and speaking.

The points raised above are in no way criticisms of the Multilink model. These points are emerging in the field and any model that attempts for a holistic accounting of bilingual language processing and cognition will have to include them as much as possible. The multilink model importantly simulates effects in which some of the earlier models had failed. It is noteworthy that Multilink considers the role of second language proficiency. Given the fact that L2 proficiency has been shown to affect both language nonselective activation (e.g., Singh & Mishra, Reference Singh and Mishra2012) and also executive control, it's a very important feature. I assume that in times to come many empirical data will be generated to test the predictions of the model for bilingual word recognition and translation.